1. Field of the Invention
This invention relates to a method of, and a system for producing enhanced output signals or signal images having certain desired properties (hereinafter referred to as signals) from an irreversible compression processor.
2. Description of the Prior Art
Many signals contain an enormous amount of information components which is desirable to compress as much as possible before recording or transmitting it. One method of signal compression that has been widely applied, is that of irreversible compression processing, such as orthogonal transform compression.
The irreversible compression may consist of the removal of parts of the information, for example, features which are of lesser visual importance, such as high frequency components of an image, and thereafter, encoding the reduced information through an entropy encoding scheme, e.g., Huffman coding. Such procedures, that are characterized by removal and reduction of information, can cause irreversible degradation in the quality of the output signal, which is proportional to the signal compression ratio.
In U.S. Pat. No. 4,807,029, there is proposed a method of reconstructing images from compression-processed image signals. The method described in this patent consists of using certain characteristics of the noise from training image signals to reconstruct the original image signals from the compression-processed signals.
There is also known an image reconstruction technique by which correction of signals can be achieved by the method of Projection Onto Convex Sets (POCS). This method consists of defining at least two sets of signals each containing specific a priori desired characteristics of enhanced signals. The intersection of these sets represents an enhanced signal having said desired characteristics. Starting from an initial signal that has one of the desired characterisitics and hence satisfies constraints of one set, by progressively approaching the intersection point of these sets, the signal is enhanced. This technique can be implemented by either a direct one step method or by iteratively alternating orthogonal projections from one set onto the other and back.
The method of the present invention may utilize the method described in the U.S. patent, or any other known per se method for producing enhanced output signals from irreversible compression signals without altering the characterising features of the signals. The proposed method achieves an enhanced output signal by using a priori information about: (i) the original input signals; (ii) any precompression encoding processing of the original input signals which may have taken place, and (iii) the characteristics of the irreversible compression processor that is employed. The a priori information can, for example, include general information, such as a restriction to a certain number of bits per pixel and/or it can be a particular property of a specific signal, such as, a particular value at a specific point of the signal.
The enhancement of signals is achieved by a POCS method or technique that increases the degree by which the signals conform to the constraints that are implied by all of the a priori information. The enhanced signals may exactly satisfy all of the constraints or only a greater degree thereof.